Dev train models
I added you to this MR so that you know what the status of the project is.
So far the most important fuctionalities have been implemented:
- prefiltering of the data based on newspaper and topic
- preprocessing as described in the previous paper
- training of MNB and BERT
- Logging of train params and
- Evaluation of train runs based on train log-files
Missing prefiltering based on the hate-speech score will be added as soon as the data is available.